Association of Coronary Artery Bypass Grafting vs Percutaneous Coronary Intervention With Memory Decline in Older Adults Undergoing Coronary Revascularization

Abstract

Importance: It is uncertain whether coronary artery bypass grafting (CABG) is associated with cognitive decline in older adults compared with a nonsurgical method of coronary revascularization (percutaneous coronary intervention [PCI]).

Objective: To compare the change in the rate of memory decline after CABG vs PCI.

Design, setting, and participants: Retrospective cohort study of community-dwelling participants in the Health and Retirement Study, who underwent CABG or PCI between 1998 and 2015 at age 65 years or older. Data were modeled for up to 5 years preceding and 10 years following revascularization or until death, drop out, or the 2016-2017 interview wave. The date of final follow-up was November 2017.

Exposures: CABG (including on and off pump) or PCI, ascertained from Medicare fee-for-service billing records.

Main outcomes and measures: The primary outcome was a summary measure of cognitive test scores and proxy cognition reports that were performed biennially in the Health and Retirement Study, referred to as memory score, normalized as a z score (ie, mean of 0, SD of 1 in a reference population of adults aged ≥72 years). Memory score was analyzed using multivariable linear mixed-effects models, with a prespecified subgroup analysis of on-pump and off-pump CABG. The minimum clinically important difference was a change of 1 SD of the population-level rate of memory decline (0.048 memory units/y).

Results: Of 1680 participants (mean age at procedure, 75 years; 41% female), 665 underwent CABG (168 off pump) and 1015 underwent PCI. In the PCI group, the mean rate of memory decline was 0.064 memory units/y (95% CI, 0.052 to 0.078) before the procedure and 0.060 memory units/y (95% CI, 0.048 to 0.071) after the procedure (within-group change, 0.004 memory units/y [95% CI, -0.010 to 0.018]). In the CABG group, the mean rate of memory decline was 0.049 memory units/y (95% CI, 0.033 to 0.065) before the procedure and 0.059 memory units/y (95% CI, 0.047 to 0.072) after the procedure (within-group change, -0.011 memory units/y [95% CI, -0.029 to 0.008]). The between-group difference-in-differences estimate for memory decline for PCI vs CABG was 0.015 memory units/y (95% CI, -0.008 to 0.038; P = .21). There was statistically significant increase in the rate of memory decline after off-pump CABG compared with after PCI (difference-in-differences: mean increase in the rate of decline of 0.046 memory units/y [95% CI, 0.008 to 0.084] after off-pump CABG), but not after on-pump CABG compared with PCI (difference-in-differences: mean slowing of decline of 0.003 memory units/y [95% CI, -0.024 to 0.031] after on-pump CABG).

Conclusions and relevance: Among older adults undergoing coronary revascularization with CABG or PCI, the type of revascularization procedure was not significantly associated with differences in the change of rate of memory decline.

Figure 1.. Cohort Creation and Follow-up

Detailed reasons for lack of participation in postprocedure trajectory modeling are included for completeness; all 1680 participants were used in modeling of the preprocedure trajectory. CABG indicates coronary artery bypass grafting; HRS, Health and Retirement Study; and PCI, percutaneous coronary intervention. ^aZero weight was assigned to participants who were not community dwelling (ie, lived in a nursing home or similar care facility) at the time of the preprocedure interview. ^bUnweighted values represent each participant contributing equally to the study sample. ^cWeighted values apply population-level sampling weights, as assigned by the HRS’s complex sampling schema, so that the study population is representative of the US population of older adults. ^dOther reasons for nonparticipation in an interview within 3 years after procedure: CABG (n = 11): 10 had no interview within 3 years, but resumed participation at a subsequent HRS interview, and 1 participated in the postprocedure interview but did not participate in cognitive testing. PCI (n = 6): 6 had no interview within 3 years, but resumed participation at a subsequent HRS interview.

Figure 2.. Adjusted Linear Mixed-Effects Models for Memory Score and Dementia Probability for CABG vs PCI Recipients

Models in A and B included 652 coronary artery bypass grafting (CABG) recipients and 1001 percutaneous coronary intervention (PCI) recipients at the preprocedure interview. Memory score is a summary measure of word recall and proxy cognition assessments roughly analogous to a population cognitive z score. Higher memory scores indicate better composite memory function. Dementia probability was derived from the memory score, the Telephone Interview for Cognitive Status, and serial-7 subtractions, and it ranges from 0 to 1.

Figure 3.. Adjusted Memory Score Trajectory Model for PCI, On-Pump CABG, and Off-Pump CABG Recipients

The model included 1001 percutaneous coronary intervention (PCI) recipients, 486 on-pump coronary artery bypass grafting (CABG) recipients, and 166 off-pump CABG recipients at the preprocedure interview. Memory score is a summary measure of word recall and proxy cognition assessments roughly analogous to a population cognitive z score (see footnote f in the Table for details). Higher memory scores indicate better composite memory function.

Figure 4.. Loess Smoother Graphs Fitting Observed Memory Score Measurements for Participants Who Met Adapted Neurocognitive Disorder Definitions

Neurocognitive disorder determinations are made on the basis of the first postprocedure memory score measurement, excluding those within 6 months of procedure, and included adapted definitions for mild neurocognitive disorder (orange; n = 267), major neurocognitive disorder (light blue; n = 88), and no neurocognitive disorder (dark blue; n = 1035). Graphs are nonparametric, unadjusted, and do not use survey weights.